Structural geology and geophysics as a support to build a hydrogeologic model of granite rock

A method developed for low-permeability fractured media was applied to understand the hydrogeology of a mine excavated in a granitic pluton. This method includes (1) identifying the main groundwater-conducting features of the medium, such as the mine, dykes, and large fractures, (2) implementing this factors as discrete elements into a three-dimensional numerical model, and (3) calibrating these factors against hydraulic data (Martínez-Landa and Carrera, 2006). A key question is how to identify preferential flow paths in the first step. Here, we propose a combination of several techniques. Structural geology, together with borehole sampling, geophysics, hydrogeochemistry, and local hydraulic tests aided in locating all structures. Integration of these data yielded a conceptual model of the site. A preliminary calibration of the model was performed against shortterm (< 1 day) pumping tests, which facilitated the characterization of some of the fractures. The hydraulic properties were then used for other fractures that, according to geophysics and structural geology, belonged to the same families. Model validity was tested by blind prediction of a longterm (4 months) large-scale (1 km) pumping test from the mine, which yielded excellent agreement with the observations. Model results confirmed the sparsely fractured nature of the pluton, which has not been subjected to glacial loading–unloading cycles and whose waters are of Na-HCO3 type.


Introduction
Low permeability fractured media have an important role in enhanced geothermal energy and waste management.Studies of this media are hampered by the contrast between conductive fractures and a nearly impervious matrix, and this heterogeneity must be considered.The characteristics and challenges of these media have led to studies in a large number of investigation sites, notably in granites, around the world: Stripa (Rouleau and Gale, 1985;Long et al., 1991), Äspö (Tsang et al., 1996;Svensson, 2001b) and Forsmark (Stephens et al., 2015;Selroos and Follin, 2013) in Sweden; Grimsel in Switzerland (Mauldon et al., 1993;Martínez-Landa and Carrera, 2005); Fanay-Augères in France (Cacas et al., 1990); Mirror Lake in New Hampshire, USA (Shapiro and Hsieh, 1991;Day-Lewis et al., 2000); Olkiluoto in Finland (Ko et al., 2015), and the Canadian Shield (Frape et al., 1984), among others.Most of these sites are located in zones affected by the last glaciation, which caused severe deformation and fracturing.In contrast, sites in southern Europe have not been subjected to glacial unloading stresses.As a re-Published by Copernicus Publications on behalf of the European Geosciences Union.
L. Martinez-Landa et al.: Structural geology and geophysics sult, most fractures have been driven by much older tectonics (Pérez-Estaún et al., 2002;Escuder-Viruete et al., 2003b) and they have been largely filled by mineral precipitates.In addition, saline water has never infiltrated these systems, so that the deep water is of Na-HCO 3 , rather than Na-Cl type.This group of sites includes El Berrocal pluton (Guimerà et al., 1995) or Ratones mine (Martínez-Landa et al., 2004), both of which are in Spain.This article is based on work conducted at the Ratones Mine in Cáceres province of Spain (Fig. 1), and contributes to improve our understanding of this type of media.
There is currently no generally accepted method for modeling low-permeability fractured media.While several different approaches have been used, they are generally combinations of methods of modeling two extremes: continuous medium and discrete fracture networks.Continuous medium fracture network-based approaches assimilate the domain to a porous medium, which includes the effect of fractures.The hydraulic conductivity field is estimated by geostatistical techniques conditioned to actual measurements of hydraulic conductivity and pressure at observation points (Neuman, 1988;Carrera and Heredia, 1988;Gómez-Hernandez et al., 1999;Day-Lewis et al., 2000;Vesselinov et al., 2001;Svensson, 2001a;Ando et al., 2003;Illman and Tartakovsky, 2005;Illman et al., 2009;Illman, 2014;Selroos and Follin, 2013;Ko et al., 2015).Discrete fracture network-based approaches represent the medium by means of fractures networks that are statistically generated.This approach is based on the assumption that fractures behave as preferential flow paths (Dershowitz, 1984;Long and Billaux, 1987;Cacas et al., 1990;Dershowitz et al., 1991;Castaing et al., 2002).More sophisticated fracture network models are channel network models, which represent only the conductive portions of the fractures planes (Moreno and Neretnieks, 1993).We have been using an intermediate mixed approach that treats the matrix and minor fractures as an equivalent, possibly heterogeneous, porous medium and deterministically simulates the hydraulically dominant fractures (Martínez-Landa and Carrera, 2006).The main drawback of this approach is that these dominant fractures must be identified and characterized.Strict characterization is only possible in intensely tested environments.Therefore, the validity of such an approach may be questionable when flow is predicted at longer scales.This is especially worrisome in view of the apparent ubiquity of scale effects.
Scale effects refer to the apparent increase of hydraulic conductivity or transmissivity as rock volume increases (Clauser, 1992;Illman and Neuman, 2000;Vesselinov et al., 2001).Multiple explanations exist for scale effects.Illman (2004) provides a thorough discussion on scale effects and their origin, pointing to other issues such as poorly developed wells (Butler and Healey, 1998) and turbulence in the boreholes (Lee and Lee, 1999).Guimerà et al. (1995) argue that long-term pumping tests are purposefully performed at the most conductive intervals, implying that they are not rep-resentative.Other authors (Sánchez-Vila et al., 1996;Meier et al., 1998;Day-Lewis et al., 2000) attribute scale effects to the connectivity among structures.Certainly, the hydraulic conductivity derived from interpreting pulse tests yields information about the closer vicinity of the borehole, which results in marked differences between values associated with intervals that intersect any conductive structure and those that are open only in the matrix (Martínez-Landa and Carrera, 2005).In cross-hole tests, pumping must be performed at intervals in high conductivity zones to maintain a significant flow rate over a long period of time, but pumping also affects points located in the matrix.If the interpretation of the tests does not take into account the existence of conductive fractures, the conductive fractures will lead to a high effective hydraulic conductivity.Martínez-Landa and Carrera (2005) reported that this effective conductivity is appropriate for predicting large-scale tunnel inflows.Such effective conductivity, however, is far larger than the average of any smallscale hydraulic test, the vast majority of which are sparsely fractured; hence, the scale effect occurs.
Thus, good representation of fractured media requires identification of the main water-conducting features.We contend that this is possible by integrating different types of information (geology, geophysics, hydrochemistry, and hydraulics).Accounting for the heterogeneity allows for the use of models for non-trivial predictions under conditions that differ from those during calibration.This has been demonstrated in previous studies (Carrera et al., 1990(Carrera et al., , 2000;;Carrera, 1993), but always at scales similar to those during calibration.Still, the best indicator of model robustness lies in its ability to predict changes in flow conditions at scales different from those during calibration.We propose that structural geology and geophysics can be used to identify large water-conducting fractures that have not been characterized by direct hydraulic tests.By assigning the hydraulic parameters of similar fracture types whose water-conducting fractures have been characterized, we could effectively extend the model scale and potentially model a large volume of rock.This article has three objectives.First, we present a method based on the above proposal and test its predictive capability.Second, we discuss the scale effects observed at the Ratones mine.Finally, we provide additional information regarding the low-permeability fractured media of the Iberian Peninsula in southern Europe.To accomplish these objectives, we analyzed field data sets from the Ratones Mine (study of the hydrogeology around an old uranium mine excavated in a granitic pluton).Data sets were obtained from geochemical, geologic-structural, geophysics, and hydrogeological studies that aided in identifying the main structures (heterogeneities), including their position, direction, dip, and extent.A three-dimensional (3-D) numerical model was then constructed, where the matrix and minor fractures were treated as an equivalent porous medium, and the identified fractures were implemented as 2-D planes embedded in the matrix.This numerical model was then used to calibrate cross-hole tests and to predict the results of long-scale pumping tests from the mine.
2 Test site

Geological and geophysical characterization
The Albalá granitic pluton is located in the southwest sector of the Iberian Massif (Central Iberian Zone of Julivert et al., 1972).The pluton is a concentrically zoned body, elongated in an N-S direction, with porphyric biotite granites in the rim and fine-grained two-mica leucogranites in the core (Fig. 1).Ratones is an abandoned uranium mine, located in the central aureoles of the pluton.The fault zone architecture in the Mina Ratones area was established on the basis of field geology, structural analysis, seismic experiments, drill cores (Ratones borehole, Sondeo Ratones, SR1 to 5) and sonic well log data (Escuder-Viruete and Perez-Estaun, 1998;Carbonell and Pérez-Estaún, 1999;Escuder-Virutete, 1999;Jurado, 1999;Pérez-Estaún, 1999;Jurado, 2000;Martí et al., 2002).Surface geology was mapped at a 1 : 1000 scale in a zone that includes the block where the seismic tomography survey was to be conducted.The resulting maps included granitic facies, dykes, ductile-brittle shears, fault zones, and granitic soil cover (lehm).The 3-D fault distribution obtained for this area is shown in Fig. 2. The post-Variscan structural evolution of the Albalá granitic pluton was established on the basis of fault kinematics and paleostress analysis in superficial outcrops (Escuder-Viruete and Perez-Estaun, 1998;Pérez-Estaún et al., 2002).This evolution includes three episodes of brittle deformations related to different stress-field configurations, which cut and reactivated ductile and ductile-brittle late-Variscan structures.
1.The first episode was extensional and produced the intrusion of Jurassic subvertical diabasic dykes, aligned following a NNE-SSW trend.The constant trend of these dykes at regional scale indicates that σ 3 was subhorizontal and WNW-ESE to NW-SE directed.
2. The second episode was characterized in the Mina Ratones area by the development of strike-slip faults with E-W direction and kilometer thickness.
The main identified structures in the Ratones area (Figs. 1  and 2) are as follows.
The North Fault (NF) has a N70 to N80 • E trend and a 55-65 • S dip, diminishing with depth to 30-40 • .
The South Fault (SF) has a N64 to N78  The damage zone is defined by small faults and kinematically related fracture sets and joints.The dykes and other fractures (as NF and SF) are hosted in an extensively fractured damage zone of hydrothermally altered granite (fractured belts).
Other relevant brittle structures of minor size include the following.
The 474 and 474' faults are two high-dip subparallel structures that trend N64 to N76 • E. Both structures connect toward the W with the 27 dyke.
The 285 brittle structure is also a sinistral strike-slip fault with a N52 to N60 • E trend and subvertical dip.Faults 474 and 285 are younger than the NF; they cut and displace it, so that it remains hydraulically disconnected.
The last deformational phase indicates that the stress state at the recent evolution was in a E-W direction.As a result, the north and south family fractures are oriented in the most favorable direction, making them ideal candidate waterconducting structures.For the purpose of subsequent hydrogeological analyses, it is important to highlight that (1) the above discussion is the result of a structural geological interpretation of 3-D seismics.(Specifically, direct seismic inversion was often modified after discussions with structural geologists, who pointed to inversion inconsistencies.Therefore, it can be stated that seismic inversion was "structural geologically biased".)(2) Some water-conducting faults (specifically 474, and 285) were identified during seismic inversion, but had not been mapped because they are located in an area covered by meadows.(3) Geophysics also helped to identify highly fractured zones (e.g., fractured granitic units, fractured belts around the dykes and the South Fault), which provided hydraulic connectivity.

Hydrogeochemical characterization
Rainwater entering into the rock-mine system is saturated with atmospheric O 2 , leading to (1) oxidation of metal sulfides present in the dykes and mineralized fractures, (2) precipitation of metal oxyhydroxides, and (3) the addition of acid to the medium.Acidic water causes dissolution of carbonates in fissure fillings, buffers the mine water pH, and promotes the precipitation of metals released by sulfide oxidation as metal carbonates (Fig. 3).In parallel, plagioclase weathering promotes kaolinite precipitation in the shallower areas.The albite becomes smectite clay, resulting in sodium bicarbonate waters in the deeper areas (400-500 m deep, borehole SR5) with transit times of tens of thousands of years (16 000 years dated with noble gases).These sodium bicarbonate waters are typical of granitic water with relatively high residence times.The groundwater circulating through granites in the Hesperian Massif is of sodium bicarbonate type, which may be significant for waste disposal in this type of rock (Gómez, 2002;Gómez et al., 2006).This is different from the chemical composition of the groundwater in granitic formations in other parts of the world, such as in the Hercinian granitic in the Chardon U mine in France (Beaucaire et al., 1999) and in the Canadian (Frape et al., 1984)   and Scandinavian (Blomqvist, 1999) shields, where chloride and sodium are the major ions.Hydrogeochemical studies help to identify water-conducting fractures.Water flowing through fractures is chemically marked by water-rock interactions along the flow path within the medium (Perez del Villar et al., 1999;Gómez et al., 1999Gómez et al., , 2001;;Gómez, 2002).This enables the identification of some fractures and potential connections.Figure 4 shows an example of the downhole geophysical loggings and records of a multiparametric device (electrical conductivity, redox potential, resistivity, gamma rays, spontaneous potential, diameter, and flowmeter) in borehole SR1, down-gradient of the mine.These data were used to identify the intersection with dyke 27, where water flows from the mine, as indicated by chemical parameters and changes in the upward velocity of flow within the borehole (flowmeter log).

Hydraulic characterization
As discussed above, geology and geophysics provide insight into the physical configuration of the fracture network.Structural geology and geochemistry help to identify which of those fractures may be water-conducting.Simple borehole hydraulic tests (pulse tests, slug tests, and constant head tests) provide the location of these structures and their hydraulic conductivity.Large-scale connectivity can be identified using cross-hole tests and hydrochemistry if the water contains a chemical tracer.The hydraulic conceptual model of the system is largely based on the hydraulic extent of the fractures and their connectivity, which is why a hydraulic testing survey was designed.Different types of measurements were used.
a. Development pumping was carried out in all SR boreholes, comprising pumping with an open borehole to withdraw all drilling materials.
b.Chemical sampling pumping was carried out of NF in borehole SR2, of SF in borehole SR4, and some intervals of SR5.Time, flow rate, and interval pressure were measured.These data were used as hydraulic test data.
Additionally, specifically designed hydraulic tests were conducted.c.Pulse, slug, and constant head single-hole tests were conducted at short intervals in boreholes SR3, SR4, and SR5.The three tests were performed between two packers separated at a constant distance and a system of pipes and valves to allow the injection or extraction of very low water volumes with high precision (Ortuño et al., 2000).
d. Three cross-hole tests were planned.For this purpose, the boreholes were divided into intervals and hydraulically isolated by packers.The position of the packers was determined to isolate the structures that had to be tested.Each interval was equipped with a pressure outlet and a water injection/extraction point.
d1.The North test had a pumping interval of S14-1 (first interval of S14 borehole, numbered from bottom to top) which was designed to characterize the dyke 27 up-gradient of the mine.
d2.The East test had a pumping interval of SR3-1, which intersects the NF, like SR2-2.After pumping for 9 days, there were no responses at any observation point.Interpretation of the seismic profiles in a subsequent stage (Pérez-Estaún, 1999;Martí et al., 2002) revealed the absence of a possible hydraulic connection between these points because the NF was cut and displaced by faults 474 and 285 (see Fig. 2).
d3.The South test had a pumping interval of SR4-1.This was located in the discharge zone of the mine, pumping within the SF.This was the only test in which all observation points reacted to the pumping.
Hydraulic characterization begins with a single-hole test (pulse, slug, and constant head).The calibrated parameters provide insight into the transmissivity field around the boreholes, and allow for identification of the most conductive intervals.Cross-hole tests were then conducted in these boreholes to identify connectivities.All hydraulic tests were interpreted using the Theis method (Theis, 1935).This model assumes that the medium is homogeneous and isotropic, thereby integrating the effect of heterogeneities into the matrix.In cross-hole tests, the preliminary interpretations were done separately for each recorded drawdown curve, which yielded a couple of transmissivity-storage (T -S) values for each observation point (Fig. 5).Note that estimated T s range over 1 order of magnitude, whereas the estimated S's, which are presumed nearly constant, range over almost 5 orders of magnitude.These highly variable storativities provide information about the connectivity between pumping and observation points through fractures.A good connection is reflected by a fast response (i.e., high diffusivity, T /S).As discussed in the introduction, the estimated T reflects largescale conductivity.Therefore, the estimated storage coefficient will be low for observation points that are well connected to the pumping interval (Meier et al., 1998).The dykes and NF and SF are accompanied by systems of minor fracturing that form part of a higher transmissivity zone situated around these structures.Points S10 and SR1-3, which display the lowest S values (Fig. 5), are connected to the pumping point through these fracturing belts.In the 3-D numerical model, this connectivity was simplified by simulating fracture planes referred to as fracture S10 to connect point S10 to the pumping interval SR4-1, and fracture SR1-3 to connect point SR1-3 to SR4-1.The remaining observation points have a lower response (i.e., higher S) to the pumping.Transmissivity and storativity are derived by fitting each borehole drawdowns, one at a time, using the Theis model.The degree of connectivity was derived from the estimates of storage coefficients.T varies from 4 to 64 × 10 −5 m 2 s −1 , while S ranges from 1 × 10 −7 to 0.43.We take S estimates to reflect connectivity.A small S (fast response) implies a good hydraulic connection between pumping and observation wells.This suggests that the best connections occur between the pumping interval and points S10 and SR1-3, while points S5 and SR1-2, and, especially mine well (point PM, Pozo Maestro), were damped by the mine influence (which behaves as a constant head boundary).Observation points S5 and SR1-2 intersect dykes 27' and 27, respectively.A priori, these should have a better response to the pumping.Drawdowns between both points were limited by the influence of the mine cavity, which acts as a constant head boundary.
3 Numerical model A 3-D numerical model was constructed to interpret the South cross-hole test, taking into account the waterconducting features.Figure 6 displays a plan view of the features introduced in the model.The granitic matrix was divided into two zones depending on the degree of characterization of the whole area.On the one hand, the internal matrix represents, in detail, all the fractures identified around the mine.On the other hand, the external matrix does not include all of the structures (because they have not all been identified).The external matrix has larger hydraulic conductivities than the internal matrix, because it includes the effect of the unidentified fractures.Its value was assigned by approximately evaluating the effective hydraulic conductivity (matrix plus fractures) of the area that was characterized in detail.The model reaches up to 600 m depth to include all measurements (the SR5 borehole reached a depth of 500 m) and all units.The shallowest layer is formed by 2-D elements and represents the lehm, which comprises altered, disaggregated, and washed granite (much like granitic sand) with high hydraulic conductivity.This layer drains rapidly in humid seasons, and carries no water in dry seasons.The altered unit reaches a depth of 20 m according to geophysics and borehole samples.The layer is highly weathered, but the granite structure is conserved.Its hydraulic conductivity and porosity are higher than those of unaltered granite due to the decomposition of feldspar.The fractured granitic unit could be identified by geophysics, and corroborated by the hydraulic characterization of a borehole 500 m deep (Fig. 7) that was drilled in the matrix far from the mine.At this depth, hydraulic conductivity is reduced, and remains almost constant up to this horizon.In this unit, the granite does not behave like the altered unit, but its fracturing index is high and, therefore, its effective hydraulic conductivity is higher than that of non-deformed granite.Finally, the non-deformed granitic unit, in contrast to the other units, has a lower fracturing index and effective hydraulic conductivity.Both the matrix effective hydraulic conductivity and specific storage decrease with depth.The upper units (lehm and altered granitic unit) are treated as separate hydraulic conductivity zones.Constant hydraulic conductivity is also adopted in the bottom portion of the model (below 350 m; Fig. 7).The linear relationship proposed by Stober (1997) was adopted between the upper and lower units.Storativity drops linearly by an order of magnitude between the surface and the bottom (600 m deep), based on the results obtained from the hydraulic tests.
The magnified image in the central part of the model (Fig. 6, right) shows the structures included explicitly in the model.Both matrix zones and fractured belts are simulated by 3-D elements; fractures are reproduced with 2-D elements (lines in the figure); finally, boreholes are introduced as 1-D elements (points in the figure).Structures included in the model (Fig. 8) through 2-D elements preserve their azimuth, dip, and interception points with the boreholes.In general, they are subvertical, except for the NF, which is slanted on the surface and cut, displaced, and tilted by fractures 474 and 285.The mine is also treated with 2-D elements, because it corresponds to the mining of part of dykes 27 and 27' (planar structures).

Model calibration of the cross-hole South test
Once the conceptual model was built and the numerical model implemented, a cross-hole test conducted downstream of the mine (South test) was calibrated.As initial parameters, those obtained from interpretation of the test with the Theis model were taken.The interpretation model of the south cross-hole test works with drawdowns; that is, initial drawdowns were zero and all prescribed fluxes, except the pumping, were set to zero.The model boundaries were placed at a distance large enough to impose a zero drawdown condition.Table 1 summarizes the results obtained alter calibrating the test.Note that fully penetrating boreholes are explicitly included in the model to represent that they connect fractures.
Figure 9 represents the resulting model fits.The pumping point is SR4-1, intercepting the SF.Observation points S10 and SR1-3 respond rapidly to the stress, because they are connected to the pumping point through small structures that form the fractured belts (simplified in the model as fracture S10 and fracture SR1-3), which intercept the SF.The remaining observation points were not that well connected with the pumping point.Pozo Maestro (PM) is the mine well, whose response (changes in pressure due to the injection test, represented as drawdowns) is to completely damp the poor connection with the pumping point.Observation points S5 and SR1-2 cut dyke 27' and 27, respectively, close to the mine.The pumping withdraws the water mainly from the shallow layer (lehm), draining it through the bed of Maderos stream, where water flows sub-superficially.Water also comes from the mine and the altered granitic unit, flowing towards the pumping point through the fractures (SF, fractured belts, and dykes).fected by each test, which ranges between a few centimeters (pulse tests) to some tens of meters (South cross-hole test).As mentioned above, the tested intervals did not coincide for all types of tests.In the case of pulse, slug, and constant head tests, they were conducted sequentially with the same interpretation, but they do not coincide with the packed off inter-   , 2006).We attribute this absence to the fact that characterization tests were performed during an unusually rainy season.The most permeable zones are located at the discharge zone of the mine, which caused some of the boreholes to become flowing wells, making it impracticable to perform a short-scale test for the highest permeable part.As a consequence, the highest values are not available for the short-scale tests.In summary, the lower values of hydraulic conductivity are associated with pulse tests for points situated within the matrix.These values are consistent with the conductivity estimated for the matrix in a 3-D heterogeneous model.The increase in hydraulic conductivity with scale is caused by the contribution of fractures to the effective conductivity obtained when interpreting the tests using homogeneous models.

Blind prediction
A 4-month pumping test from the mine was carried out with a mean flow rate of 0.0025 m 3 s −1 .All boreholes were used as observation points (Fig. 11).As for the interpretation of the south cross-hole test, all observations were introduced as drawdowns calculated from the measurements registered prior the start of the pumping (assuming steady state).The model adopted the calibrated parameters of the analysis of the South cross-hole test (shown in Table 1), and those ob- tained from the hydraulic characterization of structures not involved in the cross-hole tests.Fractures not affected by any test were assigned the transmissivities that had been obtained during calibration for fractures belonging to the same family, according to the geophysical and structural analysis.Specifically, fractures 474, and Fr4 were assigned the values obtained for fracture 285, and the SR3 dyke was assigned the transmissivity obtained for the non-mined portion of the 27 dyke.Results of the blind prediction (i.e., without calibration) are shown in Fig. 11.Overall, the fit of this blind prediction was visually excellent.Nevertheless, it becomes apparent that the estimated specific yield for the south crosshole test model was too high, which may reflect the fact that the hydraulic characterization of the zone was performed entirely during the winter and within a wet interannual cycle, so that the entire zone was fully saturated.In the prediction, it can be observed that most of the water came from the lehm, which is a very conductive layer with a relatively high specific yield (3.6 × 10 −3 ), while the altered unit specific yield was set at 4.6 × 10 −6 m −1 .This artifact reflects that water is withdrawn from the altered unit, which is the actual storage unit in dry periods.It is worth pointing out the good quality of the responses obtained at intervals SR2-1, SR2-2, SR3-1, and SR3-2, which would have displayed a much more delayed response than observed and modeled data (perhaps similar to the response observed in SR4-3 or SR3-3) if the untested (and only observed through geophysics) fractures had not been included.Perhaps more important, the slope of the late time drawdown in semi-log is quite accurate in most observation points, which suggest that the overall parameters are adequate for this test, even though it lasted some 100 times longer than the calibration tests.

Conclusions
The main objective of this work was to show that structural geology and geophysical techniques together with hydrochemical and hydraulic data can help to identify the main fractures, which carry most groundwater flow.Values of transmissivity were estimated at different scales by considering the medium as homogeneous.When plotting these values against the representative field scale, we observed a progressive increase in transmissivity with an increase in scale.This scale effect is attributed to the main fractures because all observations can be explained by incorporating such dominant fractures explicitly into the model.In fact, this model yielded an excellent fit to a 4-month long pump test that provoked responses at all observation points.The fact that the results of such a long pump test could be blindly predicted using only short-term (< 1 day) tests supports both the use of seismic geophysics and structural geology to identify the dominant fractures and their explicit incorporation into the groundwater flow model (mixed approach).The results are consistent with the hypothesis that southern European granite plutons display low hydraulic conductivity.This, together with the low aggressivity of their sodium bicarbonate groundwater, makes them appropriate for hosting nuclear waste.

Figure 4 .
Figure 4. Some of the chemical and geophysical logs recorded at borehole SR1, which helped to identify dyke 27 and fracture SR1-3.The chemical logs only display the effect of dyke 27, because all upper water (including SR1-3) is mine water.The effect of both structures can be noticed in other logs, especially in the gamma log, due to the fracture filling.Flowmeter values, recorded with an upward flow rate of pumping indicate little water flow beneath dyke 27.The same applies to a lesser extent to fracture SR1-3.

Figure 5 .
Figure5.Results from preliminary interpretation of the South cross-hole test (pumping at SR4-1).Transmissivity and storativity are derived by fitting each borehole drawdowns, one at a time, using the Theis model.The degree of connectivity was derived from the estimates of storage coefficients.T varies from 4 to 64 × 10 −5 m 2 s −1 , while S ranges from 1 × 10 −7 to 0.43.We take S estimates to reflect connectivity.A small S (fast response) implies a good hydraulic connection between pumping and observation wells.This suggests that the best connections occur between the pumping interval and points S10 and SR1-3, while points S5 and SR1-2, and, especially mine well (point PM, Pozo Maestro), were damped by the mine influence (which behaves as a constant head boundary).

Figure 6 .
Figure 6.Model geometry.The model comprises two areas.The external matrix area is treated as an equivalent porous media without explicit fractures, because not all of the main structures have been identified.At the local level (right picture), the main structures are explicitly taken into account, including the fractures, dykes 27, 27', and SR3, and the mine itself, which is excavated in both dykes.The structures are represented by means of 2-D elements.The Maderos creek is embedded in the SF.Projections of the boreholes that are closer to the mine are represented by black dots on the surface and by an arrow pointing to the borehole end.

4
Figure 10 represents the hydraulic conductivities obtained from all hydraulic tests as a function of the scale.Transmissivities derived from the tests are divided by the length of the pumped interval to convert them into conductivities.The scale of the test is determined by the rock volume af-

Figure 7 .
Figure 7. Schematic vertical section representing variations in data derived from the hydraulic characterization of borehole SR5 (500 m deep) indicate that hydraulic conductivity changes with depth.This fact was studied by Stober (1997).The model uses a modification of Stober's equation: hydraulic conductivity is kept constant up to the base of the fractured granitic unit (200 m), then changes with depth down to 350 m.From there, it remains constant down to the model bottom.

Figure 8 .
Figure 8. Main structures implemented in the model as planar structures.They are defined by 2-D elements.The model honors surface traces and dips.Downwards extension of these structures is performed with the aid of geophysics, structural geology, and intersections at boreholes.The NF is the more vertical one at its upper section, and it is cut and disconnected by faults 474 and 285 towards the south.The mine is also represented by means of 2-D elements, because it results from the exploitation of the dykes.The SF is zonated at the surface, in order to reproduce the altered zone in which the stream is embedded, where most of the water flows.

Figure 9 .
Figure 9. Results obtained after calibrating the SR4-1 cross-hole test are indicated by a black line; observed data are indicated with circles using a 3-D model.All graphs maintain the vertical scale to facilitate comparison of the results.

Table 1 .Figure 10 .
Figure 10.Permeability values obtained from interpretation of hydraulic tests performed at varying scales in different holes and intervals.In general, permeability increased with the increase in scale, except for the cross-hole test, which was performed in the highly transmissive portion of the site.See the text regarding the surprising lack of high K observation during small-scale tests.

Figure 11 .
Figure 11.Blind prediction of the long-term pumping of the mine (point PM), using calibrated with the SR4-1 cross-hole test.This pumping lasted for 4 months, and its effects reached all observation points.Measurements are represented by circles, and model results are represented by continuous line.